Process Variation Analysis
Variation Analysis explains how long completed work took and where completed cases spent their time. It supports process improvement by showing the typical result and the spread around it.
Interpret the distribution
Use the median for a typical completed case. Compare the upper quartile and maximum to find whether a minority waits much longer. A long state duration can point to a queue, unclear decision, missing information, or external dependency, but the chart alone cannot establish which cause applies.
Filter to one design before deciding a change helped. Different designs, entities, and outcomes can have different normal timing. The all-design view is useful for orientation; a single design supports improvement work.
Review cycle
- Select a design with enough completed cases.
- Find states with the longest median or widest spread.
- Inspect representative process histories for typical and long-running cases.
- Identify an actionable cause.
- Make a controlled change and compare a later completed cohort.
Monitor active cases through Dashboard and Running processes, not these completed-duration charts.
Variation Analysis compares completed process data. Choose a process design to see the time spent in each state, or leave the filter on all designs to compare them together.
Each box-and-whisker plot shows the minimum, first quartile, median, third quartile, and maximum duration in minutes. The cycle-time chart aggregates the completed state durations for each process instance, then compares those process durations by design.
The charts are based on immutable duration observations captured when a process reaches a terminal outcome. A process that is still active is not included. A small sample can look more variable than the underlying process, so use the sample count in the summary table when interpreting a plot.
Read A Box Plot
The line inside the box is the median: half of observations are below it and half are above it. The box spans the middle half of observations, from the first to the third quartile. A wide box means ordinary cases vary considerably. Long whiskers show a broader total range and may point to a smaller group of unusually fast or slow cases.
Do not interpret the maximum as the normal customer experience. Open the underlying long-running cases and look for a shared cause. A single legal hold, paused customer response, or intentionally scheduled wait can stretch the scale while most work remains stable.
Compare Like With Like
Choose comparable designs, outcomes, time periods, and operating conditions. A revision can change states or rules, and a seasonal demand period can change staffing pressure. Note these differences when presenting a comparison. Use the median and quartiles with the sample count instead of relying only on an average or the visually tallest value.
Turn Analysis Into Improvement
State a specific hypothesis, such as “cases wait in Engineering Review because required drawings arrive after assignment.” Inspect histories and evidence to test it. Then choose a response that addresses the cause: earlier information capture, clearer ownership, a due date, an escalation route, or a design revision. Continue watching active work on the dashboard while waiting for enough completed observations to evaluate the change.
Variation is not automatically waste. Some variation reflects different product complexity, risk, customer response, or required controls. The aim is to understand and manage the variation that creates avoidable delay or unpredictable outcomes.